Kubernetes Foundations

You can build an image and run it with Compose. The next thing most teams ask of you is to run it on Kubernetes — and Kubernetes does not reward guessing. This course puts a single-node cluster inside your own lab machine and has you operate a small service on it from first `kubectl get nodes` to a production handover. **The cluster and the API (lessons 1–2).** You start the control plane yourself, see how the API server, scheduler, kubelet and containerd fit together, and learn to read the cluster instead of guessing at it. You build the status service image, hand it to the cluster's own image store, and run it as a Pod in its own namespace. **Workloads and traffic (lessons 3–4).** You replace the Pod with a three-replica Deployment, roll out a release that turns out to be broken, watch what a rolling update really does when nothing is checking readiness, and roll it back — with a written change record. Then you put a Service and an Ingress in front of it and prove that traffic reaches the replicas. **Configuration and state (lessons 5–6).** Settings move into a ConfigMap and the password into a Secret; you add startup, readiness and liveness probes and watch each one act. Then you run a stateful service as a StatefulSet with a PersistentVolumeClaim, delete its Pod, and prove the data outlived it. **Operating the cluster (lessons 7–8).** Requests, limits and QoS classes; a Pod that cannot be scheduled; a container killed at its memory limit; an autoscaler that only works once it can read metrics. Then least privilege: a read-only ServiceAccount proved with `kubectl auth can-i`, and a default-deny NetworkPolicy with the few flows the namespace actually needs. **Delivery and troubleshooting (lessons 9–10).** Three planted faults to diagnose and repair in place, then Kustomize bases and overlays for two environments, with Helm and GitOps explained. **Final project.** You ship the whole stack for a synthetic logistics company to `rivermill-dev` and `rivermill-prod`, run a release and a rollback, and write the runbook the next person on call will read. Everything runs on your lab machine `linux01`, on a cluster with no internet access, so nothing depends on a cloud account. The course is a learning pathway toward backend development and platform engineering roles; it does not promise employment or any vendor certification.